Friday, December 21, 2012

The most important tool in global warming research is...

Commentor Hoo and I have been going back and forth about the observation made by the Met Office in England (a climate science center) that there really hasn't been any global warming over the past 16 years, despite large increases in atmospheric CO2.

Hoo says there has been substantial warming, almost certainly:

No, it did not stop warming. If you followed troy's link and played with the data yourself (or merely looked up the results elsewhere) then you would know that over the last 16 years the global temperature anomaly increased by 0.13 degrees plus or minus 0.13 degrees (Celsius) per decade.

What does this result mean? Statistically speaking (I know it is Greek for you), it means that the most likely value of the increase was 0.13 degrees per decade. The quoted uncertainty of 0.13 degrees per decade is two standard deviations. In plain English, the rise was between 0.065 and 0.195 degrees per decade with the probability of 68 percent, and between 0 and 0.26 degrees with the probability of 95 percent. The chance that the temperature did not increase or went down is 2.2 percent.

Let me repeat that. 2.2 percent chance that the temperature did not rise or went down.
97.8 percent chance that the temperature did rise.
68 percent chance that it rose between 0.065 and 0.195 degrees per decade.

If I were a betting man, I would not bet against global warming. If you do so then you are a fool.

2.2 percent chance that the temperature anomaly did not rise or went down.
97.8 percent chance that the temperature anomaly rose.
68 percent chance that it rose between 0.065 and 0.195 degrees per decade.

2.2 percent chance that the temperature anomaly did not rise or went down.
97.8 percent chance that the temperature anomaly rose.
68 percent chance that it rose between 0.065 and 0.195 degrees per decade.

I also calculated the trend for the period 1995 to 2009.This trend (0.12C per decade) is positive, but not significant at the 95% significance level. The positive trend is quite close to the significance level. Achieving statistical significance in scientific terms is much more likely for longer periods, and much less likely for shorter periods.
Let me unpack this for you. He is saying that the data he uses (HADCRUT) does not show a warming trend with a 95-percent confidence level, but only just so. He is not saying that there is no warming.

Go to the calculator page and select the data set he mentions: HADCRUT4 (put out jointly by the Met Office and Jones). Put the start date at 1995 and end date at 2009. Hit Calculate. You will see a rising trend with a warming of 0.140±0.164 °C/decade (2σ error bars). With a 95 percent confidence the temperature trend was between −0.02 and +0.30 degree per decade. It isn't much different from the trend seen in the GISTEMP data set I quoted earlier.

Furthermore, the line about the last 16 years is a significant distortion. It is statistically much more likely (50-to-1 odds) that global temperatures have been rising over the last 16 years than falling. The positive trend, reported by several different groups, is 1.2 degrees per century, in agreement with the rise in the last 100 years.

One more time, for those who are slow on the intake.

2.2 percent chance that the temperature anomaly did not rise or went down.
97.8 percent chance that the temperature anomaly rose.
68 percent chance that it rose between 0.065 and 0.195 degrees per decade.

Pretty impressive statistics, them.

So what does the data really look like, graphed out?

Here:

Mean global temperature 1997-2012, from Met Office, U.K. 


According to global warming mavens, it shows:
"It is statistically much more likely (50-to-1 odds) that global temperatures have been rising over the last 16 years than falling. The positive trend, reported by several different groups, is 1.2 degrees per century, in agreement with the rise in the last 100 years. [There is a] 97.8 percent chance that the temperature anomaly rose... You will see a rising trend with a warming of 0.140±0.164 °C/decade (2σ error bars). With a 95 percent confidence the temperature trend was between −0.02 and +0.30 degree per decade."

:-/


The most important tool in global warming research is a shovel.

20 comments:

  1. And the most useful tool in global warming denialism (like creationism and all the other causes fond to Egnor's heart) is cherry picking data.

    ReplyDelete
  2. Dr. Egnor,

    Which data set did you plot?

    Hoo

    ReplyDelete
    Replies
    1. You saw the link, Statistical Genius. Figure it out yourself.

      Quite a rise in temperature, eh?

      Delete
    2. Dr. Egnor,


      So it looks like you went to a secondary source, the U.K. Daily Mail. It seems like you haven't learned much from your previous experience with secondary sources.

      The data plotted in David Rose's Daly Mail article is from HadCRUT4, the data set put together by the U.K. Met Office and Phil Jones. Let's find out what the primary source (the Met Office) thinks about the secondary source (Daily Mail). In a word: misleading.

      The Met Office commented on the article here. A brief excerpt:

      An article by David Rose appears today in the Mail on Sunday under the title: ‘Global warming stopped 16 years ago, reveals Met Office report quietly released… and here is the chart to prove it’

      It is the second article Mr Rose has written which contains some misleading information, after he wrote an article earlier this year on the same theme – you see our response to that one here.


      I will let you read the response in detail, but here are my own thoughts, which aren't all that different from theirs. I am sure Dr. Troy will weigh in also.

      David Rose did not plot the data for the last 16 years (1996-2012). He chose to start his graph in August 1997, smack in the middle of a strong El Niño, when the global temperature was near its recent all-time high. Rose highlighted the very first and last points, where the temperature anomalies are the same (+0.5 degrees) and declared that the world had not warmed over that period.

      The technical term for this is cherry-picking the data.

      He also chose to plot the data without smoothing out the month-to-month fluctuations, which are of the order of 0.5 degree. You are looking for a systematic variation of less than 0.2 degrees over a decade. Such a trend is hard to see in the noise of that amplitude. This is why people plot data averaged over several months to make the trend clearer in a graph. Smoothing of course does not alter the calculated trend.

      So what happens if we crunch HadCRUT4 for the time period shown in Rose's graph? Is there no positive trend? Well, yes, there is. It is not immediately obvious in the noisy data, but the trend is +0.043±0.145 °C/decade at the 2σ level.

      The calculated trend depends on the starting point thanks to the presence of fluctuations. But it is positive:
      1995-2012: +0.109±0.119 °C/decade
      1996–2012: +0.107±0.131 °C/decade
      1997-2012: +0.058±0.136 °C/decade
      1998–2012: +0.052±0.153 °C/decade
      1999-2012: +0.095±0.162 °C/decade
      2000-2012: +0.056±0.179 °C/decade

      You can play with HadCRUT and other datasets here.

      So here is the bottom line:
      All of these intervals are consistent with a warming trend of 1 degree per century, or 0.1 degree per decade.

      And I would once again advise Dr. Egnor to stay away from unreliable secondary sources. Particularly when primary sources condemn their mishandling of the data.

      Hoo

      Delete
    3. Michael,

      I knew you didn't plot any data. Your link is to the Daily Mail website. Not the most reliable of sources (although an improvement over your blog). They don't link to the site they've taken the graph. And I can't find it on the Met Office website.

      But it's obviously a case of cherry picking. Taking too short a period. At least it didn't start with the El Niño (which would have given the false impression of a cooling) but it did end with a La Nina, one of the warmest ones on record.

      No one thinks that greenhouse gases are the only factor driving climate. The last solar cycle was weaker than usual, resulting in less solar output.

      Delete
    4. bachfiend,

      No, Rose's graph did start in the midst of a strong El Niño. See my comment above. It is cherry picking of the usual sort.

      But the most ironic part is that even with the cherry-picked start and end points, the data show a warming trend and are certainly consistent with a long-term warming trend of 1 degree per century.

      Hoo

      Delete
    5. Right, Hoo. You can "smooth" that data until it's like a baby's butt, that ain't a temperature rise. When I was a resident, we had a protocol for calculating medication doses. We would do all of the calculations, determine the proper dosage for a patient based on the parameters selected. But then, before we gave the medication, we were required to take time for a "bullshit" check. That is, we had to look at the medication, and at the patient, and see if the dose we were giving made sense. If we were giving ten vials of a medication in which a vial was the normal dose to an elderly 80 pound lady, then our calculations were screwed up.

      You need to do a commonsense check on your "calculations". There ain't no net warming in that graph. Period.

      You need to stop reading the AGW press releases, and look at the evidence objectively and with common sense.

      Delete
    6. "Taking too short a period."

      Heh. The global record shows cooling from about 1940 to 1980, with warming from 1980 to 1996, then stasis.

      The stasis-- "too short a period." is the same 15 year-length period on which global warming hysteria is based.

      Goose..., gander...

      Delete
    7. Dr. Egnor,

      It seems that you are not familiar with statistical analysis. Smoothing is averaging over several adjacent data points. It does not affect the trend.

      You could compute the average trend by hand if you knew how. The HadCRUT data set is openly available to the public. Or you can use the calculator. Smoothing or not, the trends are what they are.

      Hoo

      Delete
    8. Dr. Egnor,

      It is rather ironic that you advise me "to stop reading the AGW press releases." As this conversation showed, I work with primary data and you rely on secondary (and famously unreliable) sources. You rely on your eye, which is easily mislead, whereas I compute the trend by using standard statistical tools.

      Hoo

      Delete
    9. The trend, or lack thereof, is obvious.

      The data is smoothed, adjusted (many cold monitoring sites in Siberia went off line in the late 80's with the collapse of the Soviet Union), all of the usual global warming "science"-- lots of nature tricks and hidden declines.

      You then run dodgy data through a statistical orgone box and declare that the whole world has to yoke it's economy and governance to greenies who alone can save it.

      We've been through this before. Eugenics had lots of number crunching to "prove" that we had to sterilize defectives (the mavens of eugenics-- Galton, Pearson, Fisher, were pioneers in statistics, and "proved" their crap incessantly". Population Control loons like Erlich, Holdren etal had all sorts of "statistical" reasons why America was gonna starve and England was gonna disappear by 2000. Pesticide-phobes who banned DDT had a ton of statistics on why malarial poor people should not have DDT to help them.

      By now, prudent folks have a very finely tuned bullshit detector. You bastards have lied for a century. AGW is a lie, just like all the rest.

      Delete
    10. Michael,

      Right. I reproduced the Daily Mail graph by setting the dates 1997 to 2011 and selecting a one month moving average. It shows a warming trend of 0.0086 +/- 0.0151 degree C.

      The Earth is warming. It's not common sense to assert that there no other short term factors, such as decreased solar output, causing cooling in the same period.

      You've boasted in the past that you're ignorant of climate science. You keep on proving it.

      Delete
    11. Hoo:

      You ain't "working primarily with data". You are playing with a calculator on an AGW fansite.

      The "data" seems so freely available now. Why did the CRU crowd fight like cats getting a bath to avoid FOIA requests for data for decades, violating the law in doing so?

      Delete
    12. Dr. Egnor,

      I am not sure I understand what you are saying. Are you claiming that
      (1) the warming trend is an artifact of smoothing?
      (2) working directly with HadCRUT4 data (bypassing the calculator) will show no warming trend?
      (3) your eye is a more reliable instrument than a statistical calculator?

      Please clarify these points.

      Hoo

      Delete
    13. Michael,

      And the Daily Mail was quite happy to use the CRU data to generate their misleading graph? And leaving out the calculated trend figures which the calculator also provided, inviting anyone to fit whatever trend line he wants on such messy data? You don't want to see any warming, so that's what you see in the noisy data of the graph.

      Delete
    14. @Hoo:

      I'm saying that AGW is a con. I'm saying that the holes in your fanaticism are gaping. I'm saying that any statistical "analysis" that claims that the actual record of temperatures (shown above) for the past 16 years shows significant warming is just crap.

      I'm saying that AGW is the same kind of con that eugenics, overpopulation hysteria, anti-DDT fanaticim, and global cooling fanaticism (remember the 1970's?) were.

      You are just using science to whip up apocalyptic hysteria, for ideological and fiducial reasons.

      Today we're living through one apcalyptic prediction. We'll live through yours as well.

      We need to focus on how to make people's lives better, especially people who are in need in impoverished and repressive nations. The green agenda has been their enemy for decades.

      Delete
    15. Dr. Egnor,

      You have listed some ideological talking points, but you have not answered my questions. Let me restate them.

      Are you claiming that
      (1) the warming trend is an artifact of smoothing?
      (2) working directly with HadCRUT4 data (bypassing the calculator) will show no warming trend?
      (3) your eye is a more reliable instrument than a statistical calculator?

      Thank you in advance for supplying specific answers.

      Hoo

      Delete
    16. I'll let Dr. Curry and Dr. Trenberth answer them, in the post I'm putting up now.

      Delete
    17. Dr. Egnor,

      You keep jumping from thread to thread.

      I will comment further in the new thread. However, I ask you to answer questions (1) and (3) here, for neither Dr. Curry, nor Dr. Trenberth had anything to do with those.

      If you fail to respond, I will conclude that you do not have any good answers.

      Sincerely,

      Hoo

      Delete
    18. What is this sound I hear?
      Crickets chirping.
      Nah, can't be true.

      Hoo

      Delete